Suggestions for a folding box gone awry?

A forum just for SPCR's folding team... by request.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Feb 13, 2008 4:00 pm

What cpu voltage are you running at? You're OCd right? Go up one increment, and frankly one on memory and the NB as well.

Otherwise I'd swap out the PS to rule it out.

KansaKilla
Friend of SPCR
Posts: 381
Joined: Fri Aug 01, 2003 12:13 pm
Location: Rochester, MN

Post by KansaKilla » Wed Feb 13, 2008 4:49 pm

At 401 FSB, I was running at 1.408 in the BIOS, and with vdrop and vdroop it ended up 1.366 under CPU-Z. Currently, at 357 FSB, CPU-Z reports 1.344 but I forget what the BIOS level was. How high should I take the 401 FSB voltage? It was stable for about 28 hours under Orthos and has been stable during the week when I'm with the computer, just not over the weekends.

I'm running the Antec 380 W PSU that was in the NSK 4400.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Feb 13, 2008 5:13 pm

Right now folding SMP CPUZ reports 1.304 or 1.312volts.

:?

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Feb 13, 2008 7:25 pm

I think your power supply is a little on the weak side, yeah I understand constant draw isn't that high and all but, who's to say how steady the voltages are?

Jonnyguru.com is my site for checking on PSs.

floffe
Posts: 497
Joined: Mon May 08, 2006 4:36 am
Location: Linköping, Sweden

Post by floffe » Thu Feb 14, 2008 12:43 am

KansaKilla wrote:Others have reported that the client will stall or freeze when this happens in the middle of a WU. Any thoughts on net access and stability of the client?
Hasn't happened to me, and I've had quite a few net outages since I started folding. AFAIK it never tries to go online except when uploading/downloading a WU anyway, so I don't see why the connection should affect the calculations at all.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Thu Feb 14, 2008 7:22 am

My SMP clients crashes if network connection is lost while crunching. It's very annoying as I often leave my Quad crunching for a week unattended.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Thu Feb 14, 2008 6:46 pm

whiic wrote:My SMP clients crashes if network connection is lost while crunching. It's very annoying as I often leave my Quad crunching for a week unattended.
Yeah, I'm never comfortable leaving PCs folding, something alway goes wrong. Had 2 back-to-back WUs get stuck at the end. My Unbuntu setup seems slightly more stable from a FAH perspective.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Thu Feb 14, 2008 6:51 pm

KansaKilla wrote:At 401 FSB, I was running at 1.408 in the BIOS, and with vdrop and vdroop it ended up 1.366 under CPU-Z. Currently, at 357 FSB, CPU-Z reports 1.344 but I forget what the BIOS level was. How high should I take the 401 FSB voltage? It was stable for about 28 hours under Orthos and has been stable during the week when I'm with the computer, just not over the weekends.

I'm running the Antec 380 W PSU that was in the NSK 4400.
While both of us can run plenty stable at 333MHz FSB I think we've both seen the difference between it and native 333MHz. I'll bet the motherboard manufacturers had quite a few timings to iron out to get everything in Windows to work correctly at the higher speed. There's no other explanation for such a delay of native 333MHz. Hell even the original Biostar TForce 965 board hit 500MHz FSB fairly often.

KansaKilla
Friend of SPCR
Posts: 381
Joined: Fri Aug 01, 2003 12:13 pm
Location: Rochester, MN

Post by KansaKilla » Thu Feb 14, 2008 9:03 pm

Well, to get the voltage to not go under 1.4 while folding, I had to up the bios all the way to 1.4865 or some such. Whooda thunk there was that much drop in core voltage from bios? I've heard that Gigabyte boards have some voltage differences from what is reported in the bios to what comes out on cpu-z and the like, but dang.

I know the PSU is on the skimpy side. Just don't have the flow right now to upgrade it. Had to bargain with the boss after she freaked out when we got our first propane bill for the new house for februrary. It was more than our previous mortgage before escrow! That's why it might make sense to get a Q9450 when they come out to help with the power bill from 24-7 folding. Or 24-7 minus the weekend crashes. :wink:

At whiic, the cable internet here in KC, MO has been a little sketchy over the past 7 months. I have this suspicion that AT&T is resetting the network on the weekends to discourage P2P in the KC area. I have no proof, just a suspicion that I have. I'd be especially interested to see if the problem persists on the same machine with a different net provider. Maybe when I get back to the frozen tundra of MN in 6 weeks it will be better with Charter, but I won't hold my breath. To paraphrase Gunnery Sergeant Hartman, they are all equally worthless.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Sun Feb 17, 2008 10:34 pm

KansaKilla wrote:Well, to get the voltage to not go under 1.4 while folding, I had to up the bios all the way to 1.4865 or some such. Whooda thunk there was that much drop in core voltage from bios? I've heard that Gigabyte boards have some voltage differences from what is reported in the bios to what comes out on cpu-z and the like, but dang.
What's the big deal? You first statement implies that, only a pc folding 24/7 it will actually see 1.4865 volts for what? A total of seven minutes per day? And when my stops folding at such times temps plummet.

Vdroop is a good thing. The rebound voltages can be excessive. Frankly under load what you need to watch for is consistency, the boards ability to regulate, and not so much the value it's regulating at.

OT - Well phooey on the whole AMD business. My E6400 at 3.2GHz versus the X2 Black Edition also at 3.2GHz.

(Corrected the watts portion).
Intel versus AMD
14 minutes vs. 19 minutes - Per checkpoint on that the 1760 jobs.
150 watts vs. 135 watts - IE 10% less power for a lot less processing speed.

Both CPUs have memory running at 400MHz (DDR2-800) at stock speeds.
Last edited by aristide1 on Mon Feb 18, 2008 12:19 pm, edited 3 times in total.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Mon Feb 18, 2008 11:48 am

aristede1: "Vdroop is a good thing. The rebound voltages can be excessive."

But do vdroop reduce these voltages? To me, having read THIS, it seems the spike in voltage when current consumption rapidly either increases or decreases still exists... but rather than being a rapid increase (or decrease) followed by a rapid drop (or riset) to original vcore, the vdroop allows voltage to rise (or drop) rapidly and stay there. To me, there's still a jolt in voltage, the duration of the jolt just happens to be longer... say, for example 7 minutes instead of microseconds.

Intel says it's designed NOT to reduce voltage spikes but to limit the maximum voltage peaks to tolerable level. Why the hell couldn't that be achieved by running a (constant) lower vcore (without vdroop)? It'd save a lot of energy while computer is idling. It sure doesn't help much us overclocking folders but it'd help the rest 99% of overclocking community ... gamers for example. It'd also reduce carbon dioxide emissions if vdroop is eliminated... and vcore reduced by the same amount of course. If vcore is not dropped after vdroop mod, then the horror scenario Intel is talking might be reality, that maximum allowed voltages are exceeded.

I believe droop is implemented to:
- save cost and to
- meet Intel's specs.

You can live without vdroop if you break several Intel's specs. This includes that you not only refuse use of vdroop but also ignore VID and use a lower vcore. Motherboard manufacturers can't go against the specs or they couldn't advertise the board as Intel compatible! The overclocker has to break the specs him-/herself. Breaking vdroop is only one of the many violations... others being ignoring both VID and FID. At least we don't exceed maximum temperatures... in fact OC community, despite the higher than stock frequencies may well be running their CPUs cooler than stock.

aristede1: "Intel versus AMD
14 minutes vs. 19 minutes - Per checkpoint on that the 1760 jobs.
135 watts vs. 150 watts - IE 10% less power for a lot less processing speed."


Shouldn't that be "10% less power for a lot more processing speed" or "10% less power for a lot less processing time"? That is, if in both sentences the Intel value was mentioned before AMD's.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Mon Feb 18, 2008 12:16 pm

whiic wrote:
aristede1: "Intel versus AMD
14 minutes vs. 19 minutes - Per checkpoint on that the 1760 jobs.
135 watts vs. 150 watts - IE 10% less power for a lot less processing speed."


Shouldn't that be "10% less power for a lot more processing speed" or "10% less power for a lot less processing time"? That is, if in both sentences the Intel value was mentioned before AMD's.
My mistake here, the Intel is 150 watts, the AMD is 135. My apologies.

KansaKilla
Friend of SPCR
Posts: 381
Joined: Fri Aug 01, 2003 12:13 pm
Location: Rochester, MN

Post by KansaKilla » Mon Feb 18, 2008 7:32 pm

aristide, I guess my disgruntlement (is that a word?) is over the discrepancy in the BIOS over the voltage selected and the voltages that are reported by CPU-Z, more than anything else. It makes it harder to get in the ballpark with the voltages you expect to get when you set them in the BIOS versus what you actually get when you use monitoring programs. I should have mentioned that the voltage decrease is there at idle no matter whether I am folding or not, just not as pronounced. When folks say stuff like "don't go over 1.5 Vcore," I end up scratching my head--are they talking about 1.5 V actual or 1.5 V in the BIOS? It could be simpler. But then, that's half the fun in dinking around, right?

The folding box was up when I got back tonight. I was actually surprised. Since I have to be away for the week I was planning on dropping the OC to a level that might have been more stable so as not to lose any more points than necessary while traveling with the laptop. But since it's up, I'll see how it goes when I get back on Sunday. Hopefully, it will continue to stay up. Maybe that extra voltage helped, or maybe the cable company didn't drop the connection. Whatever, as long as it folds for SPCR (~stably).

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Mon Feb 18, 2008 8:09 pm

Kansa, other people have the same frustration with memory voltages, going to far as to trust only a voltmeter attached to known specific areas of reference.

In your case I'd keep all numbers under 1.5 volts as a precaution.

Visit Rebels Haven and some of those extreme overclocking web sites for tons of technical info, very board specific.

KansaKilla
Friend of SPCR
Posts: 381
Joined: Fri Aug 01, 2003 12:13 pm
Location: Rochester, MN

Post by KansaKilla » Mon Feb 18, 2008 10:14 pm

thanks for the site. i'll check it out.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Tue Feb 19, 2008 6:06 am

Why would it care if the NIC was down or unplugged in the middle of a WU? Sure, when a transfer needs to take place. It does check to see if the current WU is expired every once in a while though.

I do know Windows 2000 freezes up if you unplug the mouse or keyboard.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Tue Feb 19, 2008 12:02 pm

"Why would it care if the NIC was down or unplugged in the middle of a WU?"

How could I know? All I know is that it happens. If I detach network cable from the computer while it's folding, it'll continue to fold until it reaches next checkpoint, then it stops folding, no error message, no progress, CPU util and temps drop to idle values. Then it just stays there like in suspended animation. It really looks as if nothing was wrong with it as it doesn't give any type of cryptic error message.

Detaching network cable isn't the only way to make it drop to idle. Crashing network adapter is another, and I believe just changing the IP settings alone is already enough. It doesn't need to be disconnected on hardware level (and why would SMP monitor the hardware itself, instead of relying on interfaces provided by OS?)

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Tue Feb 19, 2008 5:52 pm

Really? On my C2D the network plug pops out sometimes. I've had it run to the end and then go crazy trying to send the WU.

I find Vista and Ubuntu don't care if I unplug the keyboard and mouse while they are running.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Tue Feb 19, 2008 5:55 pm

KansaKilla wrote:...(is that a word?)...
This is America, we make up words as we go along. That's what it means to live in the land of opportunity.

<Insert loud Homer Simpson burp here, to keep it real.>

KansaKilla
Friend of SPCR
Posts: 381
Joined: Fri Aug 01, 2003 12:13 pm
Location: Rochester, MN

Post by KansaKilla » Tue Feb 19, 2008 8:27 pm

Subliminabable. Simply subliminabable.

Post Reply